> ## Content Index > Fetch the complete content index at: https://marginsofcare.com/llms.txt > Use this file to discover other available public pages before exploring further. # Charité’s Launches Clé, Research Hospitals, the Cosmos Data Moat - URL: https://marginsofcare.com/charites-launches-cle-research-hospitals-the-cosmos-data-moat/ - Published: 2026-10-01T19:56:02.000Z - Updated: 2026-10-02T06:02:58.000Z - Author: Eva Worst From [ai.berlin](https://ai.berlin/news/detail/charite-switches-on-its-own-ai-platform-built-by-a-berlin-spin-off) (emphasis mine): > Charité – Universitätsmedizin Berlin has launched its own AI platform. **"Clé" is available to staff in patient care, research, teaching and administration, and runs entirely on Charité's own servers.** According to the hospital, that setup is so far unique among Germany's university hospitals. > "To our knowledge, Charité is the first university medical centre in Germany to run its own AI platform with several language and image models entirely on its own infrastructure – not as a pilot in individual departments, but across the board," says Prof. Alexander Meyer, head of the Institute for Artificial Intelligence in Medicine (IKIM) and the project's initiator… > **The platform is explicitly not a medical device. "In its currently approved form, Clé is not intended for diagnoses or treatment decisions,"** Meyer stresses. "Clinical responsibility stays with the treating professionals." AI outputs, he adds, always need to be reviewed critically. There are two things to unpack here: The first thing is that Cle is an intelligent layer that runs on top of Charite’s SAP IS-H system. The very system that will be substituted by Epic for the price of 200 million EUR. The second thing is that while Cle isn’t a medical device yet, it sure seems like it is in the roadmap. On the surface Charity’s decision seem strange: why pay so much for Epic, which comes with AI modules and whatnot, and at the same time spend resources to develop and deploy your own AI modules on a surface that is about to get disconnected? Why not just transition to the SAP-aligned and AI-first Avelios and use the money for something else? To understand this, we need to look at what Charite is as an organisation. Charite isn’t just a university hospital; it’s the [largest university hospital in Europe](https://www.charite.de/fileadmin/user%5Fupload/portal%5Frelaunch/Mediathek/publikationen/Factsheet/Charite-Factsheet%5FEN%5FKennzahlen%5F2025.pdf ), with at least 50% more beds and 50% more professors than any other university hospital in Germany. Such a massive clinical and research capacity requires the respectively massive and stable system that can support multi-campus setup in the most busy of times. Epic is already running on some of the biggest academic medical campuses across the ocean and has proven its reliability. Compared to that, Avelios is a startup that hasn’t yet had the chance to swim in the deep. Then there is the research part. [Epic comes with Cosmos](https://cosmos.epic.com/) \- a global anonymised database with hundreds of millions of patient records from across Epic’s ecosystem - that provides immediate research edge. Charite already produces around 4,000 publications per year, almost double as much as the next on the list, and the Cosmos database will only solidify and deepen that research advantage. And finally, all of this research, which is funded by public money, should materialise into real innovations and IP. This is where the Cle part comes in. University clinics have four overarching goals - teach, treat, research and translate. They work as a flywheel, each one fuelling the others, but it’s the last one that actually has the capacity to generate a positive financial return, which can then be reinvested back in the clinic. Cle is the perfect example of this: it started as a research project, got spun out into a company and is now being used to reduce costs, pending further commercialisation and revenue creation. University clinics are often the first customer of their own research projects, which also gives them advantages when it comes to costs. Not only are there internal synergies and short decision paths, but also the pilots are typically happening before the medical device certification, precisely because clinical value needs to be accessed and data generated, before a regulatory submission can be made. This way a project or a spin off can easily run a pilot for a few years without obtaining any certifications, which seems to be exactly the case. Charite will continue to do research and create innovations, some of which will be spun off while others will be absorbed internally, and that’s written in their DNA. University hospitals are [in a unique position](https://marginsofcare.com/ai-agents-on-premise-vs-in-the-cloud-part-2/) when it comes to developing and hosting their own solutions (although there are [drawbacks too](https://marginsofcare.com/ai-agents-on-premise-vs-in-the-cloud-part-2/)) but they still need a solid backbone on which they can run and if that comes with an added data moat, then all the better. If data is gold, then 310 million patient files for 200 million EUR seems like a fair price.